COPII collar defines the boundary between ER and ER exit site and does not coat cargo containers

The first sorting stations for proteins in the secretory pathway are ER exit sites (ERESs; Balch et al., 1994). These are specialized membrane domains on the surface of the ER that are identified by the COPII heterocomplex and ER-Golgi recycling proteins such as ERGIC53 (Schindler et al., 1993). The COPII protein complex has been identified in yeast and mammalian cells as a membrane coat involved in anterograde trafficking from the ER (Barlowe et al., 1994; Kirk and Ward, 2007). The complex formation is initiated by the Sar1 small GTPase, which is recruited to the ER membrane by Sec12, a protein that acts as its GDP-to-GTP exchange factor. Sar1 recruitment triggers the assembly of Sec23/24 subcomplexes with cargo proteins, and then Sec13/31 forms the outer cage of the coat (Aridor, 2018; Peotter et al., 2019; Zanetti et al., 2011). The assembling lattice deforms the membrane to become spherical or tubular membrane structures (Mancias and Goldberg, 2008; Zanetti et al., 2013). Mezzacasa and Helenius (2002) demonstrated that the COPII-labeled transitional ER defines the quality control boundary, after which misfolded proteins are not retrieved back to the ER. COPII machinery is involved in cargo recruitment, sorting, and exit out of the ER (Aridor, 2018; Barlowe and Helenius, 2016; Gomez-Navarro and Miller, 2016; Zanetti et al., 2011). However, in mammalian cells, the existence of COPII vesicles is still controversial (Mironov and Beznoussenko, 2019; Mironov et al., 2003; Zeuschner et al., 2006). Recent research focused on large cargo proteins such as procollagen stimulated alternative or additional hypotheses about the role of COPII components in cargo sorting, concentration, and export from the ER (Saito et al., 2009). Large procollagen-containing COPII carriers were described (Gorur et al., 2017; Yuan et al., 2018), but more and more data show that, alternatively, large cargo leaves the ER without COPII (McCaughey et al., 2019). TANGO, an ER resident protein, together with cTAGE5, is associated with mediating the interaction between soluble cargo proteins and COPII to facilitate their export from the ER (Saito et al., 2009; Saito et al., 2011). COPII with TANGO's help may function by forming a collar that restricts and controls access to post-ER compartments (Raote et al., 2017; Raote et al., 2018; Santos et al., 2015). Following these findings, COPII was shown to be required for stringent ER retention of misfolded and ER resident proteins (Ma et al., 2017).

Here, living intact cells expressing fluorescent protein (FP)–tagged COPII either transiently or via knock-in at endogenous levels were used to visualize ER-to-Golgi trafficking of FP-tagged cargo. We found that carrier fission, the earlier step of cargo selection, and FP-tagged cargo accumulation occur essentially downstream from COPII-labeled membranes. Confocal and immunoelectron microscopy of cells treated with Brefeldin A (BFA) and nocodazole (Noc) demonstrated that COPII defines the ER–ERES boundary, facilitating cargo selection and concentration in ERES membranes. Using mutagenesis to perturb cargo binding of COPII, we found that cargo accumulated in the ER, thus supporting COPII localization and function at the ER–ERES boundary. Furthermore, we demonstrate that a subpopulation of COPI is localized to ERESs and is also seen on ER-to-Golgi anterograde carriers. Our data suggest a revised model for COPII functioning as a stable gatekeeper at the ERES boundary rather than as a transient coat of transport carriers.

The well-established ERES distribution (Stephens et al., 2000) was confirmed using our markers expressed in living intact cells. To determine the robustness of our findings, we used COS7, HeLa, and Huh7 cell lines. The COPII subunit mCherry-Sec24C was coexpressed either with the FP-tagged ER marker reticulon or VSVG (Fig. 1 A). Fig. 1 B and Video 1 demonstrate the distribution of ERESs in living HeLa cells coexpressing COPII Sec13-mCherry and the secretory transport of VSVG-EGFP at permissive temperature of 32°C. In these cells, we also demonstrate that VSVG secretory traffic is indistinguishable from what is reported in other cell lines (Hirschberg et al., 1998). Upon temperature shift, the ER-resident VSVG is transported to the Golgi apparatus and further to the cell surface. The Sec13-mCherry is under endogenous expression as the mCherry tag was knocked in at the Sec13 C terminus using CRISPR/Cas12a (Fueller et al., 2020). The characterization of this cell line is detailed in Fig. S1. In these cells, Sec13-mCherry shows the typical intracellular distribution of ERESs and, in so doing, also serves as a control for experiments where fluorescently labeled COPII subunits (Sec13, Sec24B, Sec24C, and Sec23) were transiently expressed. In either type of experimental system, the intracellular distributions of COPII subunits appeared identical.

Next, we compared the lateral movement of membrane-associated COPII with ER-to-Golgi cargo carriers and plus-ends growth of microtubules labeled with GFP-EB3 in living intact cells. Fig. 2 A shows a maximum-intensity projection of 50 consecutive frames taken at 0.66-s intervals (see also Video 2). Compared with the rapid movement of transport carriers covering significant distances along microtubular tracks toward the Golgi apparatus, COPII-labeled membranes are practically immobile. Similar results are obtained when COPII-labeled membranes are compared with microtubule plus-end tip progression using GFP-EB3 (Fig. 2 B and Video 3). Thus, at the time frame of 3 min when carriers or microtubule plus-ends cover a distance of 10s of microns, COPII membranes are nearly stable and immobile with only minor local random movements.

Additionally, using EB3, the secretory pathway is at a steady-state compared with cells expressing VSVG, where a large mass of cargo is released from the ER. In both states, COPII-labeled membranes are stationary. These data are consistent with early observations (Hammond and Glick, 2000; Stephens et al., 2000; Westrate et al., 2020) and support the idea that COPII membranes are involved in recruiting but not transporting cargo.

As membrane-bound COPII is stable and stationary, we analyzed the early stages of ER export of VSVG. ER export was visualized at the level of a single ERES in a living intact cell. To this end, cells coexpressing the cargo protein VSVG-YFP and the COPII subunit Sec24C-mCherry were shifted to permissive temperature (32°C) after overnight incubation at the nonpermissive temperature of 39.5°C. Fig. 3 A and Video 4 show the dynamics of ER export in a single ERES in a living cell. Cargo is primarily concentrated and accumulated in growing dynamic tubular membranes that mostly do not overlap with COPII membranes and later bud as a whole.

Moreover, during multiple cycles of cargo accumulation and fission, COPII membranes are stable and stationary. Cargo accumulation and carrier fission are coupled and continuous processes. They can only be distinguished at a higher spatiotemporal resolution, as in Fig. 3 B, Video 5, Video 6, Video 7, and Video 8. Based on these image sequences, the first stage is cargo accumulation in ERES membranes (Fig. 3 B, two top panels). Subsequently, these membranes undergo fission to become membrane carriers (Fig. 3 B, two bottom panels). A key observation is that fission sites occur on contiguous ERES membranes adjacent yet segregated from COPII-coated membranes. COPII does not localize to these growing tubular membranes where the cargo is accumulating but instead remains at the tubule base without changing its intensity. An identical cargo accumulation pattern as proliferating dynamic round-shaped or tubular elements and carrier fission is apparent for other types of cargo such as the RUSH-TNF (Boncompain et al., 2012; Fig. 3 C). Also, using the knock-in Sec13-mCherry HeLa cell line transfected with VSVG-GFP (Fig. 3, D and E) and using RUSH-TNF (data not shown) show identical results.

Next, we sought to follow the transport carriers that bud from ERESs to the Golgi apparatus. The knock-in Sec13-mCherry HeLa cells were transfected with VSVG-GFP. About 10 min after the shift to permissive temperature (32°C), the VSVG-GFP within the region of interest surrounding the Golgi was photobleached with a high-power laser (Fig. 4 A) before capturing the image sequence. Fig. 4 B shows selected images and a projection demonstrating a transport carrier budding from an ERES and translocating to connect to the Golgi apparatus (see also Video 9). Next, we attempted to follow the life history of ER-to-Golgi carriers using the Rab1 molecular switch. Rab1, part of the early secretory transport machinery, was recently reported on ERESs and transport vesicles (Westrate et al., 2020). Fig. 4 C shows selected images of Huh7 cells cotransfected with VSVG-GFP and Rab1b-mCherry and incubated overnight at the nonpermissive temperature of 39.5°C. Cells were transferred to permissive temperature, and confocal images were captured at 0.8-s intervals. The path from the ERES to the Golgi apparatus of the transport carriers is also shown in the projection and Video 10. Both Rab1 and VSVG colocalization are observed on ERES and transport carriers throughout their path to the Golgi. These data demonstrate that VSVG containing ERES membranes apparently bud and move directly to the Golgi apparatus. Unlike COPII, Rab1b is bound to the carrier membrane throughout its lifetime. Furthermore, our ability to visualize the carrier membrane–bound subpopulation of the small GTPase Rab1, on the background of its soluble form, argues against the view that carrier-bound COPII is below detection levels.

Next, we asked to establish the relative localization of ERES, COPII-coated, and ER membranes. To this end, we used a previously reported experimental system where cells are simultaneously treated with BFA and Noc (Dukhovny et al., 2008). We found that the simultaneous inhibition of both Arf1 activation by BFA (Lippincott-Schwartz et al., 1989) and microtubules polymerization by Noc (Hirschberg et al., 1998) preserves the COPII-mediated cargo sorting functions within ERESs while blocking the subsequent exit of cargo from the ER (Dukhovny et al., 2008). Release of VSVG-YFP at permissive temperature in the presence of BFA and Noc results in active COPII-mediated sorting of VSVG-YFP into ERES membranes that now transform at the light microscope level into growing spherical structures (Fig. 5 A, Fig. S2, Video 11, and Video 12; Dukhovny et al., 2008; Yonemura et al., 2016). Another key feature of BFA/Noc treatment is that the COPII-cargo segregation is effectively visually emphasized. ERES membrane growth is a direct outcome of COPII-mediated cargo selection, sorting, and accumulation (Fig. S2). The localization of endogenous COPII was verified using immunofluorescence with antibodies against Sec24C (Fig. S2). The effect shown in Fig. 5 A is not a side effect of BFA and Noc because it could be reproduced using Golgicide A and colchicine, a BFA analogue with the same target, and a microtubule polymerization inhibitor (Sáenz et al., 2009), respectively (Fig. S2). A variety of cargo proteins could be localized to the ERES membrane under BFA and Noc: WT CFTR, GalT, and ERGIC53 (Fig. 5 B and Fig. S2). Interestingly, ERGIC53 under BFA/Noc treatment is restricted to the ER and is concentrated in ERES membranes indistinguishable from any other type of exported cargo. Next, localization of the membrane-bound COPII collar relative to both cargo-enriched membranes and the ER was determined using deconvolved confocal microscopy analysis of living BFA/Noc-treated cells (Fig. 5 C, Video 11, and Video 12). These data demonstrate that membrane-bound COPII is consistently positioned between the ER and ERESs. To verify whether the membrane and lumen of ERESs in BFA/Noc-treated cells are at all times continuous with those of the ER, we coexpressed mCherry with a cleavable signal sequence with VSVG-YFP. The signal sequence confers insertion into the ER lumen and its further secretion. The continuous lumen labeled with mCherry shown in Fig. 5 D and Video 13 confirms that ERESs under BFA/Noc are linked to the ER. To demonstrate that COPII can distinguish between transport-competent and -incompetent proteins in BFA/Noc-treated cells, we compared COPII-mediated sorting of WT and the Δ508 transport-incompetent mutant of CFTR. While WT CFTR accumulates in ERESs, the transport-incompetent misfolded mutant protein CFTR-Δ508 is excluded from ERESs (Fig. S2; Wang et al., 2004). These data demonstrate that COPII is functional in BFA/Noc: the membrane-bound COPII collar identifies and excludes transport-incompetent misfolded mutant proteins from entering ERESs. Moreover, cargo accumulation proceeds at a normal rate compared with untreated cells (Fig. S2; Dukhovny et al., 2008). ERESs in BFA/Noc-treated cells coexpressing the COPII subunit Sec24C and cargo were monitored using high-speed confocal microscopy to follow cycles of membrane (Fig. S2 and Video 14) and cargo accumulation. The membrane-bound COPII fluorescence levels were indifferent to the undulating cargo concentrations within ERES (Fig. S2 and Video 15) as seen for untreated cells (Fig. 3). This result is consistent with COPII stably occupying the ER–ERES boundary. At the light microscopy level, COPII forms a collar that localizes between ER and ERES membranes. To study the ultrastructure of the BFA/Noc-dilated ERESs, we performed EM using immunogold anti-GFP antibodies on cells coexpressing Sec23C-GFP and VSVG-Scarlet (Fig. 5 E). Cells used for the EM analysis were primarily examined at the light confocal level immediately after fixation (Fig. 5 E, left). The EM images of ERESs and the 3D serial sections reconstruction of the membrane structure (Fig. 5 E, bottom) reveal a highly complex topology of a hollow multilayered perforated sphere (Fig. 5 E, scheme, bottom right). These structures correspond to the membranes in which cargo is sorted and concentrated (Fig. 5 E). The scheme in Fig. 5 E, bottom right, illustrates the positioning of the section relative to the ERESs under BFA/Noc treatment. These structures were previously observed in several studies. The first set of observations reporting on analogous structures titled fenestrated spheres were from thick sectioning EM analysis of the Golgi apparatus (Rambourg and Clermont, 1992). Similar structures were also observed in BFA-treated cells and were referred to as Glumerolini (Pavelka and Ellinger, 1993). Moreover, Golgi fenestrated membranes and ERESs of BFA-treated yeast Schizosachharomyces pombe (Rambourg et al., 1995b) and Saccharomyces cerevisiae (Rambourg et al., 1995a) exhibited the same morphology and dimensional scale. The immunogold labeling of Sec23 is significantly concentrated at the base of the perforated globular membranes (Fig. 5 E, yellow arrowhead). These data underpin our observation that the membrane-bound COPII is localized between ER and ERESs.

Sec24 is a COPII subunit that directly interacts with cargo (Wendeler et al., 2007). The B isoform of Sec24 interacts with acidic export motifs found in numerous surface proteins, including VSVG and CFTR (Wang et al., 2004). To enhance the binding between Sec24B and the di-acidic motif of VSVG, we used structure-based computer modeling of Sec24B (Protein Data Bank accession no. 3EH1; Mancias and Goldberg, 2008) and designed a mutant of Sec24B by substituting valine at position 932 with the basic amino acid arginine (V932R; Fig. 6, A and B). This substitution adds another positive charge to the already existing three arginines in the binding pocket for acidic export motifs of cargo proteins. A stronger binding of the VSVG tail by the V932R mutant than the WT was predicted by both allowed docking peptide states and surface electrostatic potential analysis (Dolinsky et al., 2007; London et al., 2011; Pettersen et al., 2004). Primarily, we confirmed that the mutant Sec24B_V932R bound VSVG using coimmunoprecipitation (Fig. S3). Fluorescence recovery after photobleaching was measured for WT or mutant Sec24B in BFA/Noc-treated cells in the presence of coexpressed VSVG (Fig. 6 C). As predicted by the abovementioned models in Fig. 6, A and B, the results show that membrane-bound Sec24B_V932R has a 1.5-fold slower on/off rate and a significantly smaller mobile fraction than WT Sec24B. The on/off time scales are 4.6% × s⁻¹ and 6.9% × s⁻¹ for mutant and WT, respectively, are consistent with a stronger binding of cargo by Sec24B_V932R (Fig. 6 D)_. Expression of Sec24B_V932R blocked VSVG transport, causing its accumulation in the ER after release in permissive temperature (Fig. 6, E–G). Retention in the ER of cargo proteins in permissive temperature supports our previous results of the localization of membrane-bound COPII to the ER–ERES boundary, where it controls access of ER proteins to ERES. The phenotype of Sec24B_V932R was further studied using CFTR, another di-acidic motif–containing multi-transmembrane domain surface protein (Wang et al., 2004). Fig. 6 H shows that in BFA/Noc-treated cells, CFTR-GFP is excluded from ERESs in cells coexpressing Sec24B_V932R, but not Sec24B_WT. These data demonstrate that COPII localizes and functions at the ER–ERES boundary, essentially mediating selective cargo concentration from ER via COPII-coated membranes to ERES.

In this study, we provide data that support an alternative hypothesis for the role of the COPII machinery in ER export. Rather than COPII-coated vesicles, we demonstrate that COPII-coated membranes form a collar that defines the boundary between the ER and ERES membranes (Fig. 7 A). The traditional mechanism described for COPII recruitment, assembly, and cargo selection is compatible with our alternative model. Specifically, ER-localized Sec12-activated Sar1 recruits the Sec24-Sec23 to the membrane at the ER side of the COPII collar. Next, the Sec23-Sec24 complex interacts with transport competent cargo, and the complex translocates to release the cargo at the ERES. The drive for this COPII-cargo complex movement toward the ERES is based on two types of well-documented interactions: non-specific interaction between cargo transmembrane domain and surrounding lipids (Fig. 7 A, top right; Dukhovny et al., 2009; Sharpe et al., 2010), and specific interaction between cargo export signal and COPII (Fig. 7 A, bottom right; Kuehn et al., 1998; Lee and Goldberg, 2010). These interactions in the context of our new model are discussed in detail in the Discussion. This model is fully compatible with the new revisited hypothesis of COPII-mediated ER–Golgi direct connectivity that is gaining momentum (Griffiths, 2000; Raote and Malhotra, 2019). A subpopulation of Golgi-associated ERESs is consistently present when labeling with COPII markers (Fig. 1, A and B; and Fig. 2, A and B). ERESs are also frequently observed adjacent to Golgi cisternae in EM studies (Fan et al., 2003; Sesso et al., 1994). After the shift to permissive temperature, we visualized VSVG-YFP accumulation in Golgi membranes that appear decorated with ERES (Fig. 7 C and Video 1). This accumulation can be interpreted at least in part as mediated by direct transfer of cargo from ER to the Golgi via the Golgi-associated COPII domains. The simultaneous disappearance of the Golgi and subpopulation of Golgi-associated COPII domains upon BFA treatment (Video 16) provides some support for direct connectivity between ER and Golgi. Notably, under that BFA treatment, the rest of the peripheral non–Golgi-associated ERESs were apparently not affected (Ward et al., 2001).

The localization of COPII to the boundary between the ER and ERES membranes prompts a question about the coat protein's identity for anterograde ER-to-Golgi carriers. Several observations clearly show COPI's involvement in anterograde traffic (Scales et al., 1997; Shima et al., 1999). Both Scales et al. (1997) and Shima et al. (1999) demonstrated a sequential mode of action for COPI replacing COPII on ER-to-Golgi vesicles. COPI-positive membranes are considered as markers of the adjacent ER-Golgi intermediate compartment (Ben-Tekaya et al., 2005; Klumperman et al., 1998). However, it is still accepted in the literature that COPI is exclusively associated with retrograde transport. Confocal time-lapse microscopy of living intact LdlF cells (Presley et al., 2002) coexpressing Sec24C-mCherry and the ER marker KDEL-BFP2 demonstrated that COPI localized to Golgi apparatus membranes and Sec24C-positive ERESs (Fig. S4). Before budding, both εCOP-YFP and Sec24C-mCherry partially overlap and display local random movements as a single object on top of the ER (Video 17). Upon budding, the COPI-positive membranes rapidly move toward the Golgi. This observation challenges the retrograde role of COPI and is consistent with the idea that COPI is labeling a separate organelle such as ERGIC (Blum et al., 2000). A rapid movement toward the Golgi follows the budding of COPI membranes.

Thus, we demonstrate a new model for the function of COPII based on direct live-cell microscopy and EM. We find that COPII occupies and defines the boundary between the ER and ERESs, where it functions in cargo sorting and concentration. COPII-free cargo-loaded ERES membranes undergo fission to form ER-to-Golgi carriers.

The COPII machinery role in the concentration of cargo has been previously reported (Farhan et al., 2007; Malkus et al., 2002) and reviewed (Gomez-Navarro and Miller, 2016). Traditionally, cargo concentration is coupled to COPII-coated vesicle formation (Béthune and Wieland, 2018). However, our data demonstrate a well-defined localization of COPII to the ER–ERES boundary. Accordingly, the membranes that accumulate cargo, transform to carriers, and move to fuse with the Golgi are COPII-deficient. Our findings are consistent with several previously published studies. (i) The stationary and stable appearance of COPII-labeled domains (Hammond and Glick, 2000; Stephens et al., 2000; Westrate et al., 2020) challenges the role of COPII as a vesicle coat as it is absent from mobile carriers. Its apparent stability during cargo accumulation in adjacent ERESs is inconsistent with the expected COPII accumulation during the cargo selection-coupled coat assembly (Kuehn et al., 1998). (ii) The localization and function of COPII as a gatekeeper at the ER–ERES boundary is consistent with its emerging role in quality control (Ma et al., 2017; Mezzacasa and Helenius, 2002). (iii) Our model is compatible with the export mechanism proposed for large cargo molecules such as procollagen (Raote et al., 2020; Raote and Malhotra, 2019; Raote et al., 2018; Saito et al., 2009). Essentially, the localization of COPII to the ER–ERES boundary is preempting the need for a variety of coat configurations to fit different cargo shapes and sizes. (iv) Our model explains the observations of early recruitment of COPI to anterograde transport carriers in close proximity to the ER (Bednarek et al., 1995; Orci et al., 1997; Presley et al., 2002; Scales et al., 1997; Stephens et al., 2000). As COPII occupies the ER–ERES boundary, COPI on ERES membranes may well be involved in the fission of the final transport carrier from the ERES, also more directly explaining why BFA, an inhibitor of the ARF1 GDP-to-GTP exchange factor, inhibits ER export. Undoubtedly, the core COPII proteins Sar1, Sec13/31, Sec23/24, Sec12, and Sec16 are essential for ER export. They are involved in recruiting and concentrating cargo at ERES. It is widely accepted that upon recruitment to ER membranes, COPII coat induces buds that eventually transform to spherical 60–70-nm vesicles (Zanetti et al., 2011; Barlowe and Helenius, 2016; Gomez-Navarro and Miller, 2016; Aridor, 2018). COPII-coated vesicles were indeed observed in in vitro assays (Barlowe et al., 1994) and since then observed in other systems: in yeast (Mogelsvang et al., 2003), plants and algae (Donohoe et al., 2013; Staehelin and Kang, 2008), and mammalian cells (Zeuschner et al., 2006). However, doubts about the existence of COPII vesicles in mammalian cells persisted (Fromme and Schekman, 2005; Mironov et al., 2003; Mironov et al., 1997) and even increased over the years (McCaughey et al., 2019; Raote et al., 2017; Raote et al., 2018; Santos et al., 2015; Westrate et al., 2020). How can the COPII collar at the ER–ERES boundary act as a gatekeeper and concentrate cargo in ERES? We suggest two distinct and established driving forces that simultaneously facilitate cargo selection, entry, and concentration into ERESs. The first is based on the interaction of cargo protein with its surrounding membrane. It is widely accepted that an ER–Golgi–plasma membrane thinner-to-thicker gradient of membrane thickness is critical for targeting proteins in the secretory pathway (Sharpe et al., 2010). The transmembrane domain (TMD) of plasma membrane–destined cargo such as VSVG is longer than the average thickness of the ER membrane and thus hydrophobically mismatched with the thinner ER membrane. Alleviation of this mismatching can thermodynamically drive cargo membrane proteins’ movement from the thinner ER into the thicker ERESs (Borgese, 2016). To this end, we demonstrated that stepwise shortening of the TMD of VSVG proportionally slows its accumulation and decreases its concentration in ERESs and the trans-Golgi network, thereby slowing its passage through the secretory pathway (Dejgaard et al., 2008; Dukhovny et al., 2009). Also, the cargo receptor Erv14 in yeast was shown to preferentially bind cargo with longer TMDs (Herzig et al., 2012). The membrane tension caused by mismatching may be locally restricted to the lipids surrounding the cargo TMD. Although it may not facilitate long-range cargo movement, it can stabilize the concentrating cargo population in the thicker ERES membrane (Dukhovny et al., 2009). The second mechanism for cargo entry into ERESs is the COPII-cargo export signal–specific interaction that directly mediates the selective concentration of cargo in ERESs (Balch et al., 1994; Campbell and Schekman, 1997; Kuehn et al., 1998; Nishimura and Balch, 1997). Membrane-bound COPII exchanges regularly with a cytosolic pool (Forster et al., 2006). COPII is recruited by and subsequently replaces Sec12 and Sar1, which reside in the ER membrane (Kurokawa et al., 2016). Here, we propose that the continuous addition of COPII at the ER side of the COPII domain results in an overall ER-to-ERES directional movement of the membrane-bound COPII coat. Thus, the COPII coat binds the cargo on the ER and releases it on the other side of the COPII-coated membrane collar.

Together, these mechanisms result in the capacity of the COPII complex to generate a selective and directed flux of cargo to the ERES. Our model allows the on-demand flexible and extensive proliferation of membranes to generate large ER-to-Golgi carriers (Presley et al., 1997). The recent idea of transport in nonvesicular tubular elements (tunnels) suggested by the Malhotra and Stephens laboratories is consistent with the localization of COPII reported here (Malhotra and Erlmann, 2015; Raote and Malhotra, 2019).

We and others have found COPI colocalizing with COPII and their sequential function in ER-to-Golgi transport (Lavoie et al., 1999; Scales et al., 1997). Based on our model placing COPII at the ER–ERES boundary, we propose that COPI binds cargo carriers already on ERES membranes and not at a later stage in the ER–Golgi intermediate compartment. This hypothesis is supported by a coincidental decrease of COPII and an increase of COPI on ER-to-Golgi carriers (Scales et al., 1997). Also, at the EM level, it was shown that cargo undergoes concentration only on ERES but not later (Balch et al., 1994). The idea that the ERGIC is essentially a COPI-coated anterograde carrier that buds en bloc from ERESs is not very different from the current concept of ERGIC. The juxtaposition of COPI-positive ERGIC and ERESs was also demonstrated (Ben-Tekaya et al., 2005). In this and other studies (Farhan et al., 2008; Shima et al., 1999), COPI-positive membranes are considered to be ERGIC that is adjacent yet physically separated from ERESs.

The genetic and protein–protein interactions data surrounding COPII-mediated ER cargo export are all entirely consistent with our model. What indeed conflicts with our data is the evidence for COPII vesicles from EM data (Zeuschner et al., 2006). The conventional vesicle coat model of COPII is based on ultrastructural studies of purified COPII components and artificial membranes (Béthune and Wieland, 2018; Mancias and Goldberg, 2008). These studies could not identify the function of COPII at the ER–ERES boundary as vesicle formation was analyzed in a cell-free system that cannot assign a precise intracellular localization to the process. The use of semi-permeable cells to reconstitute COPII vesicle formation may also have altered the fragile tubular–vesicular structure of the ERESs due to potentially destructive effects of the permeabilizing detergent on membranes. The concern that a substantial amount of data on COPII vesicles is indirect has been raised previously (Mironov, 2014; Mironov et al., 2003). Recent ultrastructural studies demonstrated and analyzed COPII-mediated tubular membrane structures (O’Donnell et al., 2011; Zanetti et al., 2013). These structures observed in vitro might correlate with the COPII collar in situ. Small round COPII vesicles (Donohoe et al., 2013; Staehelin and Kang, 2008; Zeuschner et al., 2006) may be more common in lower organisms, yet are the exception rather than the rule in mammalian cells. Future work using ultrafast super-resolution microscopy could clarify this point as there should be no principal obstacle to visualize 50–100-nm vesicles labeled with fluorescent cargo or coat.

The data presented here lay the foundation for an alternative model of how, where, and when the COPII complex orchestrates the functions of the ERES domain. This model is of major significance as it allows a better understanding of protein retention mechanisms and export from the ER in heath and disease.

All reagents were purchased from Sigma-Aldrich unless otherwise stated. Human Sec24B was subcloned into pmCherry-C1 (Clontech) using SalI and BglII restriction sites and verified by sequencing. Human Sec24C was subcloned into pmCherry-C1 or pEYFP-C1 (Clontech) using XhoI and SacII restriction sites and verified by sequencing. Human Rab1B was cloned into pEGFP-C1 (Clontech) using XhoI and BamHI restriction sites and verified by sequencing. ss-mCherry (mCherry with the signal sequence from hen-egg-lysozyme). YFP-VSVGtsO45 was prepared as described elsewhere (Ward et al., 2001). GalT-YFP was prepared as described elsewhere (Zaal et al., 1999). GFP-CFTR and CFTRΔ508-GFP were a kind gift from R. Kopito (Stanford University, Stanford, CA). Rab1b was prepared as described elsewhere (Nevo-Yassaf et al., 2012). Reticulon-GFP was a kind gift from T. Rapoport (Harvard University, Boston, MA). Carboxypeptidase E was a kind gift from R. Harbesfeld (TAU, Tel-Aviv, Israel). Sec24B V932R mutant was prepared using the Quick-Change kit from Stratagene. The primer used for the PCR reaction was 5′-CAA​GAA​AAA​TTG​GGT​TTG​AAG​CTA​GAA​TGA​GAA​TAA​GGT​GTA​CTA​AAG​G-3′.

COS-7, Huh7, or HeLa cells were grown at 37°C in a 5% CO₂-humidified atmosphere. Cell cultures were maintained in DMEM supplemented with 10% (vol/vol) fetal bovine serum and penicillin and streptomycin (Biological Industries). A final concentration of 1% (vol/vol) nonessential amino acids was added to the Huh7 cells culture media. Polyethyleneimine MAX transfection reagent (Thermo Fisher Scientific) was used following the manufacturer’s protocols for plasmid DNA transfections of subconfluent COS7 and Huh7 cells. Confocal laser scanning microscopy experiments were performed 18–24 h after transfection.

Endogenous SEC13 was tagged at the C terminus with mCherry using the PCR-based method described by Fueller et al. (2020). PCR was performed on plasmid pMaCTag-P13 (Addgene; #120024), kindly provided by Marius Lemberg (Zentrum für Molekulare Biologie der Universität Heidelberg, Heidelberg, Germany), using oligos M1_SEC13, 5′-GTT​GAT​GGG​CAG​TGG​GTG​TGC​ATC​AGT​GAT​GTC​AAC​AAG​GGC​CAG​GGC​TCC​GTA​TCA​GCA​TCA​GTG​ACA​GAG​GGC​CAG​CAG​AAC​GAG​CAG​TCA​GGT​GGA​GGA​GGT​AGT​G-3′ and M2_Sec13_, 5′-CAG​GAA​GGG​GCA​GTC​CTG​GAG​CTG​GCG​GGT​GGG​GAG​CCA​GGC​CCC​ACC​TGT​CTT​GAA​AAA​AAG​CAG​AAC​GAG​CAG​TGA​CAA​ATC​TAC​AAG​AGT​AGA​AAT​TAG​CTA​GCT​GCA​TCG​GTA​CC-3′. Cells were transfected with purified PCR product, and enAsCas12a was a gift from Keith Joung and Benjamin Kleinstiver (Molecular Pathology Unit, Massachusetts General Hospital, Charlestown, MA; Addgene; #107941; Kleinstiver et al., 2019) and selected for 7 d with puromycin. After that, cells were FACS-sorted in four different fractions that were analyzed for proper mCherry localization. The fraction with the lowest intensity contained mainly cells with mCherry staining ERES. Single-cell clones were isolated and characterized. One representative clone, L6, heterozygous for the tagged insertion, was chosen for further analysis. The amino acid sequence at the fusion site is SVTEGQQNEQSGGGGSGGGGSVSKGEEDNMAIIKE (underlined, SEC13; italics, mCherry; SGGGGSGGGGS, inserted linker).

Cells were imaged in DMEM without phenol red but with supplements, including 20 mM Hepes, pH 7.4. Transfection and imaging were performed in a 35-mm glass-bottomed microwell dish (MatTek) or glass coverslips. A Zeiss LSM710 or LSM800 confocal laser-scanning microscope was used (Carl Zeiss MicroImaging). Fluorescence emissions resulting from 405-nm excitation for CFP, 488-nm excitation for GFP, and 543-nm excitation for mCherry were detected using filter sets supplied by the manufacturer. The confocal and time-lapse images were captured using a Plan-Apochromat 63× 1.4-NA objective (Carl Zeiss MicroImaging). The temperature on the microscope stage was held stable during time-lapse sessions using an electronic temperature-controlled airstream incubator. Images and videos were generated and analyzed using the Zeiss LSM Zen software and National Institutes of Health Image and ImageJ software (W. Rasband, National Institutes of Health, Bethesda, MD). High-frequency images were captured using a Nikon Ti microscope equipped with a Yokogawa CSU X-1 spinning disc, controlled by Andor IQ2 software, or a spinning disk confocal unit (Yokogawa Electric; CSU-X1) attached to an Axio Observer Z1 microscope (Carl Zeiss).

Cells cultured on coverslips were fixed at room temperature in a mixture of 2% paraformaldehyde–PBS for 10 min, then washed three times with 1× PBS + 3% FCS. Cells were permeabilized and labeled simultaneously by incubation with appropriate primary antibodies: anti-Sec24C (4 μg/ml; Abcam; catalog no. Ab122635) and 0.01% saponin at room temperature for 1 h. Cells were washed three times with 1× PBS for 5 min at room temperature and incubated for 1 h at room temperature with appropriate Cy3 secondary antibodies (dilution, 1:200; Jackson ImmunoResearch) and 0.1% saponin. Images were acquired using a Zeiss Pascal confocal laser-scanning microscope as described above or an Axio Observer Z1 fluorescence microscope (Carl Zeiss). For Sec13 detection, a rabbit anti-SEC13 antibody was used (1:1,000; R&D Systems; MAB9055-SP).

COS7 cells expressing Sec23-GFP and VSVG-Scarlet under BFA/Noc treatment were fixed 30 min after the shift to permissive temperature for immuno-EM (Beznoussenko et al., 2014) Briefly, cells were fixed with 1% glutaraldehyde in Hepes, pH 7.0, post-fixed in reduced OsO₄ for 2 h, washed, treated with 0.3% thiocarbohydrazide for 30 min, and then, after the washout, with 1% OsO₄ for 1 h (sometimes treatment with thiocarbohydrazide and OsO₄ was done three times) followed by dehydration and embedding in Epon-812. Cells were then incubated with polyclonal anti-GFP antibody (Abcam; UK AB6556) for 2 h and subsequently with anti-rabbit Fab’ fragment nanogold conjugates (Nanoprobes; #2004) for 2 h, enhanced with the GoldEnhance kit (NanoProbes) according to the manufacturer’s instruction. Epon embedding and sectioning of the gold-labeled cells, two-step EM tomography, and 3D reconstruction were performed as reported earlier (Beznoussenko et al., 2014; Beznoussenko et al., 2016). Briefly, an ultratome (Leica AFS; Leica Microsystems) was used to cut 60-nm serial thin sections and 200-nm serial semi-thick sections. Sections were collected onto 1% Formvar films adhered to slot grids. Both sides of the grids were labeled with fiduciary 10-nm colloidal gold (British Biocell International). Sections were analyzed under a Tecnai-12 electron microscope (Thermo Fisher Scientific) equipped with Analysis software.

24 h following transfection, cells were washed with PBS and solubilized in lysis buffer (50 mm Tris-HCl, pH 7.4, 1% NP-40, 0.25% sodium deoxycholate, 150 mm NaCl, and 1 mm EDTA) containing protease inhibitor cocktail (Sigma-Aldrich). Extracts were clarified by centrifugation at 12,000 g for 15 min at 4°C. Following SDS-PAGE separation, proteins were transferred onto nitrocellulose membranes and blocked with 5% low-fat milk. Membranes were incubated with specific primary antibodies, washed with PBS containing 0.001% Tween-20, and incubated with the appropriate horseradish peroxidase–conjugated secondary antibody. After washing in PBS containing 0.001% Tween-20, membranes were subjected to enhanced chemiluminescence detection analysis. For immunoprecipitation analysis, cells were solubilized in lysis buffer (see above). Cell lysates were incubated with the specific antibody for 2–3 h, at 4°C, followed by 3–18 h rotated incubation with protein A/G agarose beads (Santa Cruz Biotechnology) at 4°C. Beads were collected by slow centrifugation, washed four times with lysis buffer, and analyzed by SDS-PAGE followed by detection with a specific antibody. Immunoprecipitation from the medium was performed similarly.

Antibodies used were Sec31-mouse anti-SEC31A antibody (Becton Dickinson; 61235), actin-rabbit anti-actin (Abcam; ab8227), mCherry-rabbit anti-mCherry antibody (Abcam; 167453), VSVG-mouse anti-VSVG antibody (8G5F11; Kerafast; EB0010), GFP-rabbit anti-GFP (Invitrogen; A-11122), and tubulin-mouse anti-αtubulin (Sigma-Aldrich; T6199). Chemiluminescence was recorded with an Imager LAS4000 mini (FujiFilm Live Science USA).

Fig. S1 illustrates characterization of HeLa cells with Sec13 endogenously tagged with mCherry, using immunofluorescence microscopy and Western blot analysis. Fig. S2 shows the structure and function of ER exit sites in BFA/Noc-treated cells. Fig. S3 shows that the Sec24BV932R mutant interacts with VSVG. Fig. S4 displays colocalization to and translocation from ERESs of COPI-coated membranes. Video 1 shows a time-lapse of VSVG-EGFP trafficking in Sec13-mCherry CRSPR/CAS12 knock-in HeLa cells. Video 2 shows COS7 cell co-expressing VSVG-YFP and the COPII subunit Sec24C-mCherry after shift to permissive temperature (32°C). Video 3 shows microtubule plus-end polymerization in COS7 co-expressing EB3-GFP and the COPII subunit Sec24C-mCherry. Video 4 shows cargo accumulation and fission of carriers from ERESs in COS7 cells co-expressing COPII subunit Sec24C-mCherry and VSVG-YFP. Video 5, Video 6, Video 7, and Video 8 show cargo accumulation and carrier fission in a single ERES. Video 9 shows photobleaching of the Golgi to expose ER to Golgi carrier life history. Video 10 shows Rab1b localizes with VSVG-YFP cargo throughout the Life history of ER-to-Golgi carriers. Video 11 and Video 12 show COPII labeled membranes separate ER and ERES membranes in BFA/nocodazole treated cells. Video 13 shows that ERESs are connected to the ER in BFA/Noc treated cells. Video 14 shows cargo and membrane accumulation in a single exit site under BFA/Nocodazole treatment. Video 15 shows COPII-mediated cargo sorting in ERESs of BFA/Nocodazole-treated COS7 cells. Video 16 shows that Golgi-associated ERESs disappear upon BFA-induced Golgi membranes blink-out. Video 17 shows translocation from ERESs of COPI coated membranes in e-COP-YFP expressing LDLF cells.

Thanks to Ben Nichols (Medical Research Council), Jennifer Lippincott Schwartz (Janelia), and Nihal Altan Bonnet (National Institutes of Health) for critical reading and suggestions, Ron Kopito for the CFTR-GFP constructs, Tom Rapoport for the reticulon-GFP, Rina Harbesfeld for carboxypeptidase E–GFP, and M. Lemberg, Keith Joung, and Benjamin Kleinstiver for Addgene plasmids. Many thanks to the team of the imaging facility at the Fritz Lipmann Institute, Jena, Germany.

This work was supported by Israel Science Foundation grant 1063/16 to K. Hirschberg.

The authors declare no competing financial interests.

Author contributions: O. Shomron, I. Nevo-Yassaf, T. Aviad, I. Brodsky, Y. Yaffe, E. Erez Zahavi, Anna Dukhovny, E.H. Sklan, E. Perlson, and Y. Yonemura carried out experiments. A. Yeheskel and M. Pasmanik-Chor performed bioinformatics analysis. G.H. Patterson carried out multifocal structured illumination microscopy analysis. O. Shomron, C. Kaether, and K. Hirschberg wrote the manuscript. A. Mironov, G.V. Beznoussenko, and A.A. Mironov carried out EM experiments. M. Sannai generated endoSec13-mCherry cells.